Save the model with best results

[ryabs07]

How do we save the neural network model that yeilded best results throughout all the iterations?

For instance, the following code saves the model of final iteration which might not necessarily have been the best model:

model = dde.Model(data, net)
dde.optimizers.set_LBFGS_options(maxiter= iterations, gtol= 1e-6)
pde_resampler = dde.callbacks.PDEPointResampler(period=10)
model.compile("L-BFGS", loss_weights = [1, 10, 10000, 10000]) #
losshistory, train_state = model.train(iterations=iterations, callbacks=[pde_resampler])
dde.saveplot(losshistory, train_state, issave=True, isplot=True)
model.save("LBFGS_model.h5")

[lululxvi]

Use dde.callbacks.ModelCheckpoint https://deepxde.readthedocs.io/en/latest/modules/deepxde.html#deepxde.callbacks.ModelCheckpoint